A&A 433, L49-L52 (2005)
M. J. Harris1 - J. Knödlseder1 - P. Jean1 - E. Cisana2 - R. Diehl3 - G. G. Lichti3 - J.-P. Roques1 - S. Schanne4 - G. Weidenspointner1
1 - Centre d'Étude Spatiale des Rayonnements, BP 4346, 31028 Toulouse Cedex 4, France
2 - IASF, via E. Bassini 15, 20133 Milano, Italy
3 - Max-Planck-Institut für extraterrestrische Physik, Postfach 1603, 85740 Garching, Germany
4 - DSM/DAPNIA/SAp, CEA Saclay, 91191 Gif-sur-Yvette, France
Received 1 February 2005 / Accepted 23 February 2005
It is believed that core-collapse supernovae (CCSN), occurring at a rate once per century, have seeded the interstellar medium with long-lived radioisotopes such as 60Fe (half-life 1.5 Myr), which can be detected by the -rays emitted when they -decay. Here we report the detection of the 60Fe decay lines at 1173 keV and 1333 keV with fluxes 3.7 1.1 per line, in spectra taken by the SPI spectrometer on board INTEGRAL during its first year. The same analysis applied to the 1809 keV line of 26Al yielded a line flux ratio 60Fe/26Al = 0.11 0.03. This supports the hypothesis that there is an extra source of 26Al in addition to CCSN.
Key words: ISM: abundances - nucleosynthesis - gamma-rays: observations
The radioactive isotopes 26Al and 60Fe are both believed to be produced in massive stars that end their lives as core collapse supernovae (CCSN; masses ). Further, they have similar half-lives ( yr and yr respectively) which are much longer than the characteristic interval between CCSN (100 yr). Therefore they will accumulate in the interstellar medium until a steady state is reached, and indeed this steady-state abundance of 26Al has been detected via a flux in the 1809 keV line from its -decay (Mahoney et al. 1982; Diehl 2001).
The sky distribution of 26Al has been mapped (Knödlseder et al. 1999) and, as expected, the line emission is dominated by the Galactic plane, where massive stars are found. It might be expected that the distribution of 60Fe line emission (at 58 keV, 1173 keV and 1333 keV) would be very similar. However (as we will see in Sect. 4) there are subtle differences in the sources of the two isotopes (mass and metallicity of star, depth within star and effect of the final explosion) which make the relative distributions of 60Fe and 26Al a potential source of information on fine details of massive-star evolution.
In this paper we report the detection of two of the 60Fe lines by the SPI instrument, part of the INTEGRAL mission. The significance of this detection is not sufficient for us to draw any conclusions about the sky distribution relative to 26Al. However the mission is expected to continue for several years, by which time there may be enough data for spatial information to be extracted. Earlier measurements (summarized by Harris et al. 1997) did not detect 60Fe, and yielded only upper limits on the 60Fe/26Al ratio, until the preliminary detection reported by RHESSI (Smith 2004), which is consistent with ours (Sect. 3.4).
The INTEGRAL spacecraft was launched October 17, 2002 into a high-inclination, high-eccentricity orbit with a 3-day period. It carries two major -ray instruments, the co-aligned IBIS and SPI. Although each performs both imaging and spectroscopic tasks, IBIS is designed for fine spatial resolution while SPI has superior energy resolution. Its 19 Ge detectors achieve 0.3% resolution around 1 MeV; imaging at the level within a field is enabled by a coded mask permitting differential illumination of the detectors as a function of angle. In our analysis we do not make use of this capability, because of the weakness of the lines.
One input into our analysis is therefore an assumption about the Galactic distribution of 60Fe. We used the distribution of far-infrared (240 m) emission mapped by COBE/DIRBE (Hauser et al. 1998) which is expected to be a good guide to the distribution of massive stars and is in fact one of the best predictors of 26Al emission (Knödlseder et al. 1999). The observations available for use in our analysis were made during the first year of operations (orbits 19-130) and largely came from two core (instrument team) programmes - two deep exposures (4 Ms) to the Galactic centre, and a periodic scan along the Galactic plane. We also made use of data publicly released as of 2004 December. This amounted to 13.5 Ms distributed over the whole sky, but favouring the inner Galaxy locations where the DIRBE 240 m and COMPTEL 26Al maps and the massive star population peak (see e.g. the exposure map of Knödlseder et al. 2005).
During orbits 19-130 SPI's performance was nominal, with all 19 detectors operating at full efficiency (effective area 70 ) and resolution (2.5 keV FWHM at 1.3 MeV). The energy resolution was maintained against the degradation caused by cosmic-ray impacts by in-orbit annealing, which was effective in restoring performance. In our analysis we allowed for the fact that comparisons between spectra taken at different epochs would have found slightly different widths and energies for the same line, due to this time variability of resolution (Sect. 3.2). Similarly, we analyzed separately single-detector events (SE) and "multiple'' events (ME) where the -ray energy was deposited in two detectors because their effective spectral resolution and instrumental backgrounds differ somewhat.
The basic principle behind our analysis was the creation of models describing the time variation of the background count rates, which dominate the cosmic signal by a factor 100. The models were made up from physically plausible environmental quantities that ought to contribute to the background. Empirically, the prompt component of the background (both line and continuum) is well described by the count rates in the Ge detectors when saturated (energy loss >8 MeV, hereafter "GEDSAT'' rates). But the prompt interactions also create radioactive isotopes with finite half lives, which are strong sources of background lines. The most important example is radioactive 60Co, which decays with a 7.6-year mean life emitting two -ray lines identical with those from cosmic 60Fe. We modelled this time series by convolving the source term (GEDSAT) with an exponential increasing on a 7.6 yr time scale. We define the term template to mean a function of environmental variables whose time series we will fit to the background data, whether simple (like GEDSAT) or complex (such as GEDSAT 7.6 yr). If there is a cosmic source of the lines, it should follow a quite different time series (SPI's successive exposures to the 240 m map), so we make this the third term in the fit.
The best results were obtained when all three components (two background templates and the expected 60Fe flux) were fitted simultaneously to the count rates by detector and pointing in 1 keV bins (Sect. 3.1). We performed an alternative analysis in which the "off-pointings'' which are empty of 60Fe according to the 240 m map were treated separately (Sect. 3.2). The results of this are used only as a check, and are not included in our final result, since they contain several systematic errors. Finally, since the cosmic 60Fe/26Al ratio is perhaps the most interesting quantity which we can derive, we have analysed the 1.809 MeV line of 26Al in the identical way, so as to eliminate systematics from the ratio if possible (Sect. 3.3).
In this analysis the count rates in the 60Fe lines were measured as described below in each of the 6821 pointings in our data set, and fitted by the templates GEDSAT and GEDSAT 7.6 yr, and by the expected cosmic 60Fe flux (i.e. SPI's exposure to the 240 m map). A template for the continuum which consisted of GEDSAT convolved with the time series of count rates summed over two bands on either side of each line (1163-1169, 1177-1183 keV, and 1318-1328, 1349-1359 keV) was kept fixed in the fit. The line count rates to be fitted were obtained by summing over the intervening intervals (avoiding blending lines as in 1335-1348 keV). The line count rates in 1 keV bins, and the best-fitting amplitudes of GEDSAT and GEDSAT 7.6 yr, are shown for SE data in Fig. 1. Clearly the variation of the count rates away from the line is explained by high amplitudes of the GEDSAT variable, while those in the line itself are strongly influenced by the 7.6 yr time-scale in the other template, which otherwise has amplitude zero as expected.
|Figure 1: Amplitudes of the GEDSAT background term (full line) and the GEDSAT 7.6 yr background term (dot-dash line) fitted (along with the 60Fe map exposure) to the time series of SE count rates for 1161-1184 keV. Data points - background counts in 1 keV bins with flat continuum subtracted.|
|Open with DEXTER|
The method thus explains the background (typically of the total) well, and we plot the amplitudes of the 60Fe term in the fit (signal-to-noise ) to get our result (Figs. 2-3). The spectra appear to be quite free of systematics, except for the strong 69Ge L-shell-capture decay line at 1337 keV. A more subtle systematic error (Sect. 2) might arise in the line energies and widths in Figs. 2 and 3 from the variation of the energy resolution (due to degradation and annealing) during the period covered by the measurements, which we consider below (Sect. 3.2c and footnote 1). With this caveat we find the mean strength of each of the 60Fe lines to be 3.7 1.1 by fitting the spectra in Figs. 2 and 3 by a model consisting of either one or two Gaussian lines of fixed width 2.4 keV FWHM plus a flat continuum. The significance is best visualized by summing all four lines together (Fig. 4).
|Figure 2: Spectra of a) SE and b) ME from the analysis of Sect. 3.1, with fitted 60Fe lines of strength 3.4 2.5 and 5.6 2.7 respectively.|
|Open with DEXTER|
|Figure 3: Spectra of a) SE and b) ME obtained from the analysis of Sect. 3.1, with fitted 60Fe lines of strength 2.3 1.6 and 6.7 3.2 respectively. The blending 69Ge decay line is also fitted.|
|Open with DEXTER|
In this alternative analysis method we attempted to derive a "universal'' combination of templates which describes the variation of the 60Fe count rates during those pointings for which we are fairly confident that there is no signal, i.e. "off''-pointings towards Galactic latitudes > . Only background templates were fitted to these data. To the optimum combined template we applied a correction for the discrepancies in energy resolution due to detector degradation effects, which must exist between the off-pointings and the Galactic pointings, using the algorithm described by Knödlseder et al. (2004).
|Figure 4: Spectra of Figs. 2 and 3 overlaid and summed to give the total 60Fe signal. Dotted line - the 3.7 line deduced from combining Figs. 2a,b and 3a,b.|
|Open with DEXTER|
|Figure 5: Spectrum of SE obtained by the analysis of Sect. 3.2. Full line - fitted 60Fe line of strength 7.9 2.1 . Dashed line - fitted 62Ni line of strength 4.0 1.9 (1163 keV), and inferred line of equal strength at 1173 keV.|
|Open with DEXTER|
We could then obtain a measurement of the 60Fe lines by fitting the fixed combination of templates and expected cosmic line strengths to the Galactic pointing data. An example of the results is shown in Fig. 5. There are clearly systematic effects due to the failure of the template to remove the time series of blending lines with various half-lives, notably 62Ni (1173 keV, prompt), 52Mn (1334 keV, d) and 69Ge (1337 keV, d). Even the inclusion in the combined template of convolutions of these lifetimes with GEDSAT did not remove these lines. We made the following corrections for such systematic errors:
A measurement of the 1.809 26Al line by exactly the same method as in Sect. 3.1 yielded a result 3.4 0.2 , for a 60Fe/26Al line flux ratio of 0.11 0.03, corresponding to an abundance ratio 0.23 0.08.
The close agreement between the results of the two analyses (Sects. 3.1, 3.2) suggests that the correction to the line widths and energies for the varying instrumental resolution probably has little effect in Sect. 3.1. In view of this lack of systematic errors, we regard this as our best result, i.e. 3.7 1.1 , with a possible systematic error 0.3 10-5 due to the non-uniform energy resolution.
The significance is rather better than that of Smith's (2004) result, 3.6 1.4 , but the agreement between the two is very good.
In the context of massive star evolution, 26Al comes from zones containing free protons (H and Ne burning) while 60Fe requires a substantial free neutron abundance (C, Ne and to a small extent He burning). In the final supernova explosion, they will be produced in roughly equal amounts (Limongi & Chieffi 2003). Prantzos (2004) pointed out the contradiction between this expected CCSN ratio and results such as ours and Smith's (2004), where the ratio is 0.2. His conjecture that there is a large additional source of 26Al which acts prior to the core collapse and explosion appears to be borne out by the models of Palacios et al. (2005), who find this source to be the massive winds expelled during the Wolf-Rayet (WR) phase. The key point is that there is a large abundance of 26Al in H-burning layers which are close enough to the surface for the wind to expel it during the star's short presupernova life. The 60Fe abundance is much further inside.
Surprisingly, therefore, the Galactic distributions of the 26Al and 60Fe lines may be quite different. WR stars differ from the average SNII progenitor in being (a) somewhat more massive on average and (b) highly dependent on metallicity. The 26Al map exhibits "hot spots'' in areas like Cygnus which are too young for even their most massive stars to have become SNII, but in which WR winds are already active (Knödlseder et al. 2002); 60Fe emission should not be seen from these regions. The metallicity gradient in the Galaxy is substantial enough for excess 26Al emission to be seen from the inner Galaxy in COMPTEL data (Palacios et al. 2005); 60Fe should be more evenly distributed. When the possibility is factored in that some 60Fe is made in SNIa (Iwamoto et al. 1999), and some 26Al in AGB stars and novae (Diehl 2001), which have a quite different history and distribution, it appears that we must expect the unexpected when the data become sufficient for a 60Fe map to be made.
The SPI project was completed under the responsibility and leadership of CNES. We are grateful to ASI, CEA, CNES, DLR, ESA, INTA, NASA and OSC for support.